# A Systematic Comparison between Extractive Self-Explanations and Human Rationales in Text Classification

arXiv: https://arxiv.org/abs/2410.03296 · HTML: https://mlsanity.com/paper/2410.03296/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 20 of its 38 references that are in the map.
- First posted on arXiv 4 Oct 2024.

## Builds on (most cited first, 20 of 20)

- [Learning Important Features Through Propagating Activation Differences](https://mlsanity.com/paper/1704.02685/) (2017)
- [Towards Faithfully Interpretable NLP Systems: How should we define and evaluate faithfulness?](https://mlsanity.com/paper/2004.03685/) (2020)
- [ERASER: A Benchmark to Evaluate Rationalized NLP Models](https://mlsanity.com/paper/1911.03429/) (2019)
- [Attribution Patching Outperforms Automated Circuit Discovery](https://mlsanity.com/paper/2310.10348/) (2023)
- [CLIMATE-FEVER: A Dataset for Verification of Real-World Climate Claims](https://mlsanity.com/paper/2012.00614/) (2020)
- [The Unreliability of Explanations in Few-shot Prompting for Textual Reasoning](https://mlsanity.com/paper/2205.03401/) (2022)
- [Reframing Human-AI Collaboration for Generating Free-Text Explanations](https://mlsanity.com/paper/2112.08674/) (2021)
- [AttnLRP: Attention-Aware Layer-Wise Relevance Propagation for Transformers](https://mlsanity.com/paper/2402.05602/) (2024)
- [Are self-explanations from Large Language Models faithful?](https://mlsanity.com/paper/2401.07927/) (2024)
- [XAI for Transformers: Better Explanations through Conservative Propagation](https://mlsanity.com/paper/2202.07304/) (2022)
- [Can Large Language Models Explain Themselves? A Study of LLM-Generated Self-Explanations](https://mlsanity.com/paper/2310.11207/) (2023)
- [Inseq: An Interpretability Toolkit for Sequence Generation Models](https://mlsanity.com/paper/2302.13942/) (2023)
- [On Measuring Faithfulness or Self-consistency of Natural Language Explanations](https://mlsanity.com/paper/2311.07466/) (2023)
- [MambaLRP: Explaining Selective State Space Sequence Models](https://mlsanity.com/paper/2406.07592/) (2024)
- [Properties and Challenges of LLM-Generated Explanations](https://mlsanity.com/paper/2402.10532/) (2024)
- [Evaluating the Interpretability of Generative Models by Interactive Reconstruction](https://mlsanity.com/paper/2102.01264/) (2021)
- [Being Right for Whose Right Reasons?](https://mlsanity.com/paper/2306.00639/) (2023)
- [Rather a Nurse than a Physician -- Contrastive Explanations under Investigation](https://mlsanity.com/paper/2310.11906/) (2023)
- [RaFoLa: A Rationale-Annotated Corpus for Detecting Indicators of Forced Labour](https://mlsanity.com/paper/2205.02684/) (2022)
- [Exploring the Trade-off Between Model Performance and Explanation Plausibility of Text Classifiers Using Human Rationales](https://mlsanity.com/paper/2404.03098/) (2024)

## Built on by (most cited first, 1 of 1)

- [Not All Explanations Simulate Equally: Comparing Verbalized Feature Attributions and Self-Generated Rationales](https://mlsanity.com/paper/2606.01148/) (2026)

## Abstract (arXiv)

Instruction-tuned LLMs are able to provide textitan explanation about their output to users by generating self-explanations, without requiring the application of complex interpretability techniques. In this paper, we analyse whether this ability results in a textitgood explanation. We evaluate self-explanations in the form of input rationales with respect to their plausibility to humans. We study three text classification tasks: sentiment classification, forced labour detection and claim verification. We include Danish and Italian translations of the sentiment classification task and compare self-explanations to human annotations. For this, we collected human rationale annotations for Climate-Fever, a claim verification dataset. We furthermore evaluate the faithfulness of human and self-explanation rationales with respect to correct model predictions, and extend the study by incorporating post-hoc attribution-based explanations. We analyse four open-weight LLMs and find that alignment between self-explanations and human rationales highly depends on text length and task complexity. Nevertheless, self-explanations yield faithful subsets of token-level rationales, whereas post-hoc attribution methods tend to emphasize structural and formatting tokens, reflecting fundamentally different explanation strategies.